25 Best AI Prompts for GST Professionals

These 25 AI prompts cover the full GST workflow an Indian CA or GST professional runs every month — reconciliation, IMS triage, return preparation, notice handling, classification questions, and client communication. Each prompt is written to reduce hallucination risk by forcing Indian context and blocking the model from inventing law it does not actually know. Anonymise client facts before you paste anything, and verify every citation and figure afterward — the discipline covered in How to Write Better AI Prompts for Accounting and Tax Work and What AI Gets Wrong About Indian Tax.

Reconciliation and ITC prompts (1–5)

1. "Act as a GST reconciliation reviewer. Given this anonymised list of GSTR-2B exceptions [describe categories, e.g. missing-in-2B, value mismatch], draft a triage order — which category to chase first and why, in terms of ITC at risk versus effort. No invented rule numbers."
2. "Draft a supplier follow-up email template for invoices showing as missing in GSTR-2B for the current period. Keep it factual and non-accusatory; leave the invoice list as a placeholder table."
3. "Outline the difference between a Rule 37 (180-day) reversal check and a Rule 37A (supplier non-payment) reversal check as an internal training note for a new junior. Mark exact deadlines as [VERIFY CURRENT RULE]."
4. "Given this anonymised description of a recurring ITC mismatch pattern [describe], list possible root causes a CA should investigate before concluding it is a supplier filing gap. Do not conclude which cause it is."
5. "Turn this verified reconciliation summary into a one-page client memo explaining the net ITC position for the month. Do not alter any figure; flag any sentence that still needs my confirmation."

For the mechanics behind these prompts, see How AI Can Automate GSTR-2B vs Purchase Register Reconciliation.

IMS prompts (6–8)

6. "Draft an SOP for a junior handling the Invoice Management System dashboard monthly — what to auto-accept, what to flag for review, and what must never be left to deem-accept. No invented portal features."
7. "Summarise this anonymised list of IMS-flagged duplicate invoices into a short list of questions for the client's accounts team. Neutral tone."
8. "Create a one-page explainer for a client on what IMS accept/reject/pending means for their ITC, in plain English. Do not state specific deadlines — mark them as [VERIFY]."

Return preparation prompts (9–13)

9. "Draft a pre-filing GSTR-3B checklist for a multi-GSTIN trading client on the QRMP scheme. Separate data-entry checks from judgement calls. No fabricated notification numbers."
10. "Outline the typical causes of a GSTR-1 vs GSTR-3B turnover mismatch for a services client, without concluding which cause applies to my anonymised facts: [describe]."
11. "Draft an internal reviewer's checklist for reverse-charge liability (RCM) entries before a GSTR-3B is filed, covering common omissions like import of services and GTA. Leave rates as [VERIFY]."
12. "Turn this verified GSTR-9 variance list into a client-facing explanation of what each variance means and what action, if any, is proposed. Do not change any number."
13. "Create a data-collection checklist for GSTR-9C preparation, listing what should come from the books versus what should come from the filed returns. No invented turnover thresholds — mark as [VERIFY CURRENT LIMIT]."

Deeper workflow context: How AI Can Help with GST Compliance and Return Preparation.

Notices and litigation prompts (14–17)

14. "Summarise this anonymised GST scrutiny notice, para by para, into a table of allegation and requested document. Add no legal analysis or citations."
15. "Draft a reply skeleton — headings only — for a notice alleging an ITC mismatch between GSTR-3B and GSTR-2B. Leave every legal ground and section reference blank for me to complete."
16. "Draft a standard covering letter and request for personal hearing for a GST show-cause notice reply. Professional tone. No section numbers, no case law, no figures."
17. "Given this anonymised list of issues raised across several client notices this quarter [describe], group them by theme so I can decide which pattern needs a firm-wide client advisory."

Full drafting and analysis workflow: Using AI to Analyse GST Notices and Identify Key Issues and How AI Can Help Draft Replies to GST Notices.

Classification, rate, and RCM prompts (18–21)

18. "Given this anonymised description of goods or services [describe], list the factors that usually distinguish HSN/SAC classification for similar items. Do not state a final rate or code — mark as [VERIFY CURRENT NOTIFICATION]."
19. "List the questions a CA should ask a client before deciding whether a supply is a composite supply or a mixed supply, based on these anonymised facts [describe]. Do not conclude."
20. "Draft a checklist for identifying reverse-charge liability on import of services for a services company, listing common trigger transactions. Leave the current rate as [VERIFY]."
21. "Outline the place-of-supply questions relevant to an anonymised inter-state services transaction [describe]. List only the questions, not the conclusion."

Client communication and firm ops prompts (22–25)

22. "Rewrite this partner's rough note into a polite client email requesting pending GST documents for the month. No legal threats, professional tone."
23. "Draft a monthly GST status update template for a client with three GSTINs, covering filing status, reconciliation status, and open exceptions. Leave figures as placeholders."
24. "Create a training quiz (with answers) for new joiners on what client GST data must never be pasted into a consumer AI chat."
25. "Summarise our firm's approved-tools policy for GST AI use into a five-line reminder for the internal WhatsApp group."

How to use this list without creating new risk

• Anonymise before you paste — GSTIN, legal name, director names, and precise figures should become placeholders.
• Keep the constraint line — every prompt above that touches law includes a "do not invent" or "[VERIFY]" instruction; do not delete it to save a sentence.
• Name a reviewer — for anything that leaves the firm, someone other than the drafter checks citations and figures before it goes out.
• File the trail — save which prompt, which draft, and who reviewed it, the same working-paper discipline you would apply to a junior's output.

Frequently asked questions

Are these prompts safe to use with real client GST data?

Only after anonymisation. Replace GSTIN, legal name, and any identifying figures with placeholders before pasting into a consumer chat. If a task genuinely needs real ledger or portal data, use a domain tool built for client data rather than a general chat window.

Will these prompts stop AI from inventing GST notification numbers?

No prompt eliminates the risk entirely. Each prompt below includes a constraint asking the model to mark uncertain law as [VERIFY] rather than invent it, which reduces how often a fake notification number reaches a client document — but every citation still needs to be opened and checked.

Do these prompts replace a GST reconciliation tool like GSTAgent?

No. These prompts are for drafting, checklists, and explanation structure. Actual GSTR-2B matching against your purchase register needs a tool connected to your books and the GST Portal, such as GSTAgent, not a chat prompt.

How should a firm organise a prompt library like this one?

Group prompts by workflow stage — reconciliation, IMS, return prep, notices, classification, client communication — the same way this list is organised, and add an anonymisation reminder and a named reviewer field to each entry so juniors know who checks the output before it leaves the firm.

The takeaway

A working GST prompt library is less about clever phrasing and more about consistent guardrails: anonymised facts, a named deliverable, and an explicit instruction not to invent law. Use the 25 prompts above as a starting library for your firm, adapt the bracketed placeholders to your own client mix, and keep the verification step non-negotiable regardless of how confident the output sounds. For research-grade citations rather than drafting help, pair this list with a domain tool such as TaxBotGPT or Vaive.